Abstract
When identifying a face the majority of humans initially look at a point of fixation (just below the eyes) that optimizes perceptual performance as predicted by a theoretical foveated ideal observer (Peterson & Eckstein, 2012). Here, we use a simulated scotoma paradigm to investigate the potential effects of macular degeneration on human optimal points of fixation (OPF) in face identification and compare these to the predictions of a foveated ideal observer with a central scotoma (S-FIO). We also evaluate observers' ability to adapt to the scotoma and learn new initial fixation strategies to optimize recognition performance. Methods: Seven observers completed a 1 of 10 face (15 deg) identification task in luminance noise with a gaze contingent display which simulated a central scotoma (radius = 8 deg). In the first study, observers made free eye movements in eight alternating blocks (1000 trials) with face viewing times of 350ms and 1500ms. In the second study, we assessed the OPFs by having observers fixate 1 out of 4 horizontally centered positions on the face (~375 trials for forehead, eyes, nose, and mouth in random order) for 350ms. Finally, observers repeated the free eye movement study. Results: The human forced fixation study showed that the simulated scotoma shifted the OPF downwards toward the tip of the nose (p < 0.05) which was predicted by the S-FIO model. However, surprisingly, observers failed to change their eye movement strategy (first and third studies) to initially fixate the new scotoma-induced optimal point of fixation. Conclusions: Our findings show unlike object following and search tasks (Kwon et al, 2013 ), humans have difficulty adapting to the scotoma and re-learning optimal fixations for face identification. Results show the potential use of the S-FIO as a benchmark to evaluate changes in OPFs for humans with low vision disorders.
Meeting abstract presented at VSS 2015